Does ventilator-associated event surveillance detect ventilator-associated pneumonia in intensive care units? A systematic review and meta-analysis
In our systematic review, the pooled VAE prevalence among patients who received mechanical ventilation in the ICU, 13.8 %, was higher than the observed 11.9 % pooled prevalence of VAP. This result is reasonable, because the VAE paradigm aims to identify a broader range of ventilator-associated complications, including both infectious complications (such as pneumonia, tracheitis, and tracheobronchitis) and non-infectious complications (such as atelectasis, pulmonary embolism, pulmonary oedema, and ventilator-induced lung injury) [36]. In a previous study VAP was estimated to be the most common complication, accounting for about 25–40 % of VAE, followed by fluid overload at 20–40 %, ARDS at 10–20 %, and atelectasis at 10–15 % [37]. Theoretically, by excluding non-infectious complications among VAE, IVAC should be more representative of VAP, and its prevalence should be closer to but still higher than VAP. However, in our meta-analysis, the pooled prevalence of IVAC was lower than that of VAP, and even in the sensitivity analysis, both VAC and IVAC were lower than VAP after limiting the evaluation to studies that used stricter diagnostic criteria. This result indicates that VAE surveillance might miss a certain number of cases of VAP.
Actually, among the 11 studies included in our meta-analysis that reported both VAC and VAP [19–22, 24–26, 28, 31, 32, 34], the pooled sensitivity of VAC for the detection of VAP was not satisfactory. Only 41.8 % of cases of VAP could be identified by using the VAC criteria; in other words, VAC surveillance missed about 60 % of ventilated patients who developed pneumonia. Similarly, the pooled PPV of VAC from nine studies [19, 21, 22, 24–26, 28, 32, 34] also indicated a poor capability for VAP detection. Only 23.2 % of patients who met the VAC criteria would be diagnosed as having VAP. Even excluding the non-infectious complications among VAE, the pooled sensitivity and PPV of IVAC, possible VAP and probable VAP for the detection of VAP were still low.
Most patients with VAP did not meet the VAE criteria, mainly because they did not meet the requirements for stable baseline mechanical ventilator settings or threshold levels of worsening gas exchange. In a study by Lilly et al., 70.8 % of patients with VAP did not have 2 days of stable oxygenation in the time frame required by the VAE criteria [26]. Moreover, this percentage was 82.6 % in a study by Annop at al., and only 17.4 % for insufficient levels of worsening gas exchange [35]. In contrast, in the study by Klouwenberg at al. [28], among VAP episodes that did not fulfil the criteria for VAE, those with no baseline period of stability accounted for only 24.0 % of the cohort, and those with insufficient increase in ventilator settings accounted for 76.0 % of the cohort. Similarly, in a study by Stoeppel et al., these rates were 39.4 % for insufficient period of stability followed by worsening oxygenation and 47.5 % for duration of respiratory deterioration less than 2 days [24].
Another explanation for the poor validity of VAE criteria for identifying VAP is that VAE criteria do not rely on chest radiography, which is the most sensitive indicator of pathologically diagnosed VAP [38, 39]. On the other hand, although the established diagnostic criteria for VAP are widely clinically accepted and applied, recent VAP criteria are seriously flawed in the subjectivity of clinical diagnosis, which might be another factor affecting the consistency of detection results between the two surveillance methods [40].
In our meta-analysis, although there were high rates of pooled specificity and high NPV for the VAE paradigm, the value of VAE for VAP detection was limited, given that negative screens are caused not only by cases in which the entity is absent, but also by those in which the entity is missed or not clear. In fact the VAP diagnostic criteria, which have been regarded as the gold standard for screening tests, are not objective or specific criteria, so high specificity or NPV for VAE does not indicate that an acceptable proportion of cases of VAP were detected by the screening test.
Some characteristics of the populations identified by VAE and VAP surveillance also significantly differed. Both VAE and VAP could prolong the length of ventilation and hospital stay, but the risk intensity was different in these two paradigms. By definition, VAE and VAP should differ in ventilation duration, because VAP requires the patient to receive mechanical ventilation for more than 2 days, while at least 4 days are required for VAE. However, interestingly, the pooled WMD of ventilation duration for VAE was about ?3 days compared with VAP. This result implies that patients with VAE who did not meet VAP criteria such as fluid overload, ARDS, and atelectasis, tended to receive a shorter duration of mechanical ventilation. Similarly, the mean hospital LOS for VAC and IVAC was about 4–6 days shorter than that of VAP. These differences may have been due to discrepancies in the severity of comorbidities and differences in the timing of extubation.
Additionally, the in-hospital mortality of VAC and IVAC was approximately twofold higher than that of VAP. In fact, a sustained decrease in oxygenation is an independent risk factor for mortality in ventilated patients [41], and thus, higher mortality associated with VAE may due to the VAE criteria aiding in the detection of more severe patients with poorer oxygenation [20]. Only patients with complications severe enough to merit the threshold levels of worsening gas exchange met the VAE criteria, whereas patients with slight worsening of gas exchange could still be diagnosed with VAP [34]. Indeed, about 60 % of patients diagnosed with VAP did not meet the VAE criteria in our meta-analysis.
Overall, we found that patients’ duration of ventilation and hospital stay were shorter in the VAE paradigm than in the VAP paradigm, while in-hospital mortality was higher in the VAE paradigm than in the VAP paradigm. In other words, the characteristics of patients identified by VAP surveillance were not accurately reflected by VAE surveillance. Confounding complications in VAE cases could have an influence on the significance of risk factors. For example, ARDS was the most common complication in VAE (46.8 %) in the study of Lilly et al. [26] while it only accounted for 16.4 % of patients with VAE in the study of Boyer et al. [21]. Consequently, the OR of in-hospital death for VAE compared with VAP in the former study was 1.50 (95 % CI 0.88?~?2.56) with no statistical difference, whereas it was 2.29 (95 % CI 1.19?~?4.43) reflecting a statistical difference in the latter study. The difference in distributions of complications may affect the population characteristics of VAE surveillance, which aims to identify a broader spectrum of complications of mechanical ventilation.
VAE surveillance has several advantages. First, VAE diagnosis is less time-consuming than traditional VAP diagnosis. The VAE paradigm was designed to rely on objective measures that can be easily assessed by professionals in the detection of infection. A study conducted in two hospitals indicated that VAE reviewers required 12 h to manually diagnose 400 ventilated patients, while the traditional VAP reviewer required 260 hours [34]. Furthermore, objective measures can be easily coded into computerized programmes. Consequently, software-based automatic data collection processes can further reduce the time needed for VAE identification. In a previous study, automatic VAE surveillance required only 1 minute to assess 110 patients, compared to 60.7 minutes using manual surveillance [42].
Second, the VAE paradigm maximizes the objectivity of surveillance to improve comparability [10]. Quantitative measures of VAE are commonly available in every ICU; thus, the objective criteria enable different institutions to compare their rates with greater confidence, such that differences in rates reflect differences in patients and processes of care rather than subjective and unquantifiable surveillance biases.
Third, including a broader spectrum of complications is beneficial to identifying a population of patients with serious complications who have not been acknowledged previously [2]. In fact, the VAE paradigm is able to identify not only patients with complications of mechanical ventilation but also those with severe respiratory compromise or progressive underlying disease, despite optimal care. Therefore, the broader spectrum is beneficial to monitoring critically ill patients in the ICU, making it possible to prevent severe complications at an early stage.
Strengths of our study include the quantitative methodology of the systematic review, a large sample size for estimating the prevalence of each type of VAE, and the assessment of consistency between VAE and VAP surveillance. However, our meta-analysis also has limitations. First, heterogeneity is a common problem for meta-analyses of observational studies, particularly those that involve proportions [43–45]. We attempted to explain heterogeneity by performing subgroup analyses, but after an exploration of the factors that were likely to contribute to the variation, such as study design and population characteristics, the heterogeneity remained unexplained. Although objective measures were used as the criteria for VAE monitoring, they were implemented independently in different hospitals and ICUs. Therefore, it is difficult to ensure that all surveillance activities are homogeneous in clinical practice. These constraints and variations in setting, such as patients’ baseline characteristics, data collection methods, and surveillance systems, may account at least in part for the significant heterogeneity observed [28].
Second, not every study reported prevalence or the original number of cases of VAE within each ventilation duration group, which is why the number of included studies within each ventilation duration sub-group was not equal, particularly for the small number of studies in groups with MV ?3 days, MV ?4 days, and MV ?5 days. The insufficient number of studies included in these groups led to an unstable pooled prevalence with a broader 95 % CI in the meta-analysis; even 95 % CIs in which the lower limit was negative were observed in several groups in our study. Therefore, the results in these groups should be interpreted with caution, and more high-quality studies with standardized ventilation duration groups are needed in the future.
Third, we did not conduct sub-group analysis in different types of ICU because of insufficient sub-group data within each type. The type of ICU type be a significant confounding factor affecting the prevalence and population characteristics of patients with VAE and VAP [26, 30]. Although the studies included in the meta-analysis reported their own ICU type, most provided the overall data from mixed ICU types rather than the sub-group data for each type. Consequently, it was difficult to separate total data into sub-group data according to ICU type in most included studies.
Finally, not all studies in our meta-analysis used the same diagnostic criteria. We included three studies that did not strictly meet CDC/NHSN criteria for VAE. A multicentre study in France slightly adapted the VAE definition, taking into account the change in PaO2/FiO2 with regard to the level of PEEP as a more reliable criterion for the assessment of worsening oxygenation [22]. Two studies used an early definition of VAE prior to the implementation of VAE surveillance by the NHSN in January 2013 [33, 34]. These early studies diagnosed VAE using looser criteria relative to the CDC/NHSN definition. In addition, our meta-analysis contains five studies that applied stricter VAP diagnostic criteria with quantitative culture of specimens [19, 21, 24, 28, 35]. After limiting the meta-analysis to these studies using stricter diagnostic criteria, the pooled estimates of each VAE type and of VAP changed, but these changes have limited effects on the relationship between VAE and VAP: the VAE paradigm still missed a certain number of cases of VAP, in-hospital mortality was still higher in patients with VAE than with VAP, and ventilation duration was still shorter in patients with VAE than with VAP.